BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Australia/Melbourne
X-LIC-LOCATION:Australia/Melbourne
BEGIN:DAYLIGHT
TZOFFSETFROM:+1000
TZOFFSETTO:+1100
TZNAME:AEDT
DTSTART:19721003T020000
RRULE:FREQ=YEARLY;BYMONTH=4;BYDAY=1SU
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:19721003T020000
TZOFFSETFROM:+1100
TZOFFSETTO:+1000
TZNAME:AEST
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260114T163640Z
LOCATION:Meeting Room C4.9+C4.10\, Level 4 (Convention Centre)
DTSTART;TZID=Australia/Melbourne:20231214T115000
DTEND;TZID=Australia/Melbourne:20231214T120000
UID:siggraphasia_SIGGRAPH Asia 2023_sess170_papers_291@linklings.com
SUMMARY:Single-Image 3D Human Digitization with Shape-guided Diffusion
DESCRIPTION:Badour AlBahar (Kuwait University); Shunsuke Saito, Hung-Yu Ts
 eng, Changil Kim, and Johannes Kopf (Meta); and Jia-Bin Huang (University 
 of Maryland)\n\nWe present an approach to generate a 360-degree view of a 
 person with a consistent, high-resolution appearance from a single input i
 mage. NeRF and its variants typically require videos or images from differ
 ent viewpoints. Most existing approaches taking monocular input either rel
 y on ground-truth 3D scans for supervision or lack 3D consistency. While r
 ecent 3D generative models show promise of 3D consistent human digitizatio
 n, these approaches do not generalize well to diverse clothing appearances
 , and the results lack photorealism. Unlike existing work, we utilize high
 -capacity 2D diffusion models pretrained for general image synthesis tasks
  as an appearance prior of clothed humans. To achieve better 3D consistenc
 y while retaining the input identity, we progressively synthesize multiple
  views of the human in the input image by inpainting missing regions with 
 shape-guided diffusion conditioned on silhouette and surface normal. We th
 en fuse these synthesized multi-view images via inverse rendering to obtai
 n a fully textured high-resolution 3D mesh of the given person. Experiment
 s show that our approach outperforms prior methods and achieves photoreali
 stic 360-degree synthesis of a wide range of clothed humans with complex t
 extures from a single image.\n\nRegistration Category: Full Access\n\nSess
 ion Chair: Xiangyu Xu (Xi'an Jiaotong University)\n\n
URL:https://asia.siggraph.org/2023/full-program?id=papers_291&sess=sess170
END:VEVENT
END:VCALENDAR